Skip to content

mhenr18/injector

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

27 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

injector

injector provides a way to inject arbitrary .dylibs into a running process. Additionally, it provides the injected code with access to the standard I/O associated with the injector, even if the target process is sandboxed.

injector is known to work on OSX 10.9.

Building

Building is done using Make, with no configuration step. Note that there's no installation step, so the entire build process is performed with a single make invocation. This will build a universal injector that can inject code into both i386 and x86_64 binaries.

Usage

Because injector uses task_for_pid(), it will need to be run with root privledges (i.e using sudo).

Invocation is as follows (likely with a sudo preceding):

injector target_pid payload_path [payload_args...]

where target_pid is the PID of the process you're targeting and payload_path is a path to a .dylib file containing your payload. The .dylib may be universal (but must at least contain code with the same architecture as the target).

Your payload .dylib must contain a payload_entry function, whose signature is as follows:

void payload_entry(int argc, char **argv, FILE *in, FILE *out, FILE *err);

This function is called on a new thread upon injection. in, out and err are files that correspond to the stdin, stdout and stderr of the injector. Don't close them in your payload.

The injector will run as long as the out and err files are kept open (which effectively means until payload_entry returns).

Implementation

injector is built on Jonathan Rentzsch's mach_inject, which does the heavy lifting of actually injecting code into other processes. Every time the injector is used, it generates a session UUID. A piece of bootstrap code is injected into the target and has the session UUID passed to it. This bootstrap code runs as a new thread and so does not block the target's execution. That session UUID is passed through to the payload as its first argument.

The bootstrap code does the following:

  • Allocates thread-local storage.
  • Creates a new pthread that we can use to continue bootstrapping.
  • Suspends the bootstrap thread and starts the pthread.
  • Creates three named fifos in the temporary directory of the process. (which have the session UUID in their names).
  • Loads the payload .dylib from the same directory (named using the session UUID).
  • Invokes the payload's payload_entry function.

By using a session UUID, it's possible for the injector binary to watch the file system event stream and get notified when the payload has created the fifos. This allows us to find a safe place to copy the payload to.

We can't use a predetermined location and just pass that through with the bootstrap injection code, because that location might not be within the sandbox of the target process. By having the process effectively tell us where it can read and write, we sidestep the sandboxing issue.

Once notified of the fifo locations, the injector copies the payload into the same directory as the fifos and sets up redirection of the standard files to them.

About

Code injection + payload communications for OSX (incl. sandboxed apps)

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published